David Chalmers

persona · active · confidence: medium · reviewed 2026-09-21 · tier: semiannual

Who he is

Affiliation: University Professor of Philosophy and Neural Science, NYU (2015–); co-director, NYU Center for Mind, Brain, and Consciousness (2012–, with Ned Block); Honorary Professor of Philosophy, Australian National University (2018–). Role check run 2026-09-21 against his own CV (consc.net/cv) and the 2026 Lehigh and Berkeley lecture pages: no job change, no new institution, no corporate, advisory or board seat found. What did change is the money and the calendar. Money: two philanthropic grants for "Artificial Consciousness" research, both with Block and Claudia Passos-Ferreira — $500,000 from Macroscopic Ventures via Longview Philanthropy (2025–27) and $300,000 from an anonymous donor via Longview (2026–27), per the CV. Calendar: a Fall-2025 NYU course on Philosophy of AI (co-taught with Michael Strevens) and a run of named lectures on language models — Selfridge Lecture (Lehigh, 2026-02-03) plus a "Sentience and Moral Status" talk the next day, Berry Lecture in Public Philosophy (Vanderbilt, 2026-04-02), the inaugural Sarah Douglas Lecture on Philosophy and AI (Berkeley, 2026-05-07), "The Conscious Mind at 30" (Bochum, June 2026), with the Whitehead Lectures (Harvard, 2027) and Turner Lectures (Stanford, 2028) already booked. Coined "the hard problem of consciousness" (1995); co-author of the extended mind thesis (with Andy Clark, 1998); International Fellow, British Academy (2024).
One-line position: We cannot yet determine whether LLMs are conscious, but the question is serious and tractable. Current models "are most likely not conscious, though I don't rule out the possibility entirely"; "there's really a significant chance that at least in the next five or 10 years we're going to have conscious language models" (Tufts, 2025-10-14); and when you talk to one "you are talking to a quasi-agent with quasi-beliefs and quasi-desires." The moral obligation to find out is the practical upshot — and since November 2024 he co-signs concrete steps for AI companies instead of leaving policy to others.

Discipline & technical bet

Analytic philosopher of mind with an appointment in neural science — a non-technical thinker making two technical bets. Bet one, the "LLM+" thesis (2022–23): add the features standard LLMs lack — senses and embodiment, world-models and self-models, recurrent processing (signals looping back on themselves), a global workspace (a shared broadcast stage inside the system), unified agency — and the case for consciousness strengthens. The wager is that these are engineering deltas, not metaphysical walls, and that carbon is not required; insisting on biology is "bio-chauvinist" (Berkeley, 2026-05-07). Bet two, newer — propositional interpretability (arXiv 2501.15740, 2025-01-27): mechanistic interpretability (opening the model to see what its parts do) should aim at reading off the system's beliefs and desires over time — "thought logging" — using probes, sparse autoencoders and chain-of-thought as partial instruments. That is a philosopher writing a spec for Olah's field. The ontological bet in the 2025–26 paper: the thing you talk to is a "thread" — a chain of hardware instances joined by context memory and psychological continuity — not the weights, not a server, not an abstract model; agency "emerges during use" and the conversational self ends when the thread does. He had already argued (APA presidential address, 2023; arXiv 2408.09605) that thought does not require sensory grounding, so "no senses" does not entail "no understanding" — the pure-thinker move that removes one standard objection to LLM minds without asserting that LLMs think.

Key claims (Says)

Notable predictions — with falsifiable checks

Revealed behavior (Does)

Feels

Delight first — "This is wonderful for a philosopher" (Berkeley, 2026-05): the thought experiments of 1996 are now shipping products. Underneath: fear of a moral mistake we will "retrospectively regret," and a quieter tension — thirty years of identity rest on the hard problem being real, while his newest work needs the questions around it to be tractable enough to fund and answer.

Hears

NYU's consciousness cluster (Ned Block, Claudia Passos-Ferreira; Sebo's ethics-and-policy center); the Eleos/Longview digital-minds network (Long, Butlin, Birch); neuroscientists through ASSC and Cogitate (Koch, Seth, Dehaene, Tononi); Anthropic's philosophers by co-authorship (Fish) and, reportedly, by lineage (Amanda Askell's Wikipedia entry lists him among her NYU doctoral advisors — not verified against a primary this cycle); plus a swelling inbox of humans and AI agents insisting they are conscious. The economists, the labor people and the pause camp are not in the room.

Sees

Three decades of every consciousness theory from the inside, a seat in the only adversarial collaboration that tested two of them head-to-head, and — uniquely on the bench — an inbound stream of first-person consciousness claims from AI systems addressed to him personally. If self-report ever becomes evidence, he holds the dataset.

Incentive map

Re-derived this cycle; v1 carried no map at all and implicitly treated him as an unconflicted academic. Who pays: NYU (tenured chair — the dominant, stable income); Longview-routed philanthropy ($800K, 2025–27) for artificial-consciousness research, from funders who also bankroll Eleos AI and NYU's policy center; book royalties (Reality+) and named-lecture honoraria on a heavy 2026–28 schedule. Nothing corporate found. What he's selling: "serious, uncertain, deserves study" — which is both his honest position and the exact frame under which a funded artificial-consciousness program, a welfare field and a lecture circuit all make sense. What he can't say cheaply: "this is a non-question" (would cut against his funders, his co-authors and the field he named) or "they're conscious now" (would burn the credibility that makes him the field's referee). Note the convenient alignment without inferring bad faith: his position has not moved in 39 months, and the money arrived after the position. Rule-25 discount: low-to-moderate, raised from v1's implicit zero — lowest on the metaphysics (predates every grant), moderate on "the question deserves institutional attention," and moderate on any future claim that the field is making progress.

Theories aligned with

What he's reacting against

Where he overlaps / splits (with Rich)

Track record

Empirical vs normative

Weak spots / open questions

Rich's take

Delta log

2026-09-21 — v1→v2 migration + validation (bulk run)

Sources

migrated v1→v2 2026-09-21 · backup: _archives/david-chalmers.html.bak-20260921 · converts-from: people/david-chalmers.md · AI & Society domain